An Implementation of Grouping Nodes in Wireless Sensor Network Based on Distance by Using k-Means Clustering

Wireless Sensor Network (WSN) is a network consisting of several sensor nodes that communicate with each other and work together to collect data from the surrounding environment. One of the WSN problems is the limited available power. Therefore, nodes on WSN need to communicate by using a cluster-based routing protocol. To solve this, the researchers propose a node grouping based on distance by using k-means clustering with a hardware implementation. Cluster formation and member node selection are performed based on the nearest device of the sensor node to the cluster head. The k-means algorithm utilizes Euclidean distance as the main grouping nodes parameter obtained from the conversion of the Received Signal Strength Indication (RSSI) into the distance estimation between nodes. RSSI as the parameter of nearest neighbor nodes uses lognormal shadowing channel modeling method that can be used to get the path loss exponent in an observation area. The estimated distance in the observation area has 27.9% error. The average time required for grouping is 58.54 s. Meanwhile, the average time used to retrieve coordinate data on each cluster to the database is 45.54 s. In the system, the most time-consuming process is the PAN ID change process with an average time of 14.20 s for each change of PAN ID. The grouping nodes in WSN using k-means clustering algorithm can improve the power efficiency by 6.5%.